High Tech Prognostication for Ovarian Cancer Patients
High Tech Prognostication for Ovarian Cancer Patients
Abstract & Commentary
Synopsis: Gene expression profiles predict early relapse in ovarian cancer after platinum-paclitaxel chemotherapy.
Source: Hartmann LC, et al. Clin Cancer Res. 2005;11:2149-2155.
Given the widely variant survival characteristics of women with advanced ovarian cancer, discrete quantifiable differences in tumor biology must exist. Heretofore, accurate characterization has been elusive. Hartmann and colleagues approached this problem by studying gene expression profiles in a homogenous cohort of advanced and previously untreated ovarian cancer patients (n = 79), with the intent of developing a predictive model which could discriminate early treatment failures (< 21 months) from those with prolonged survival (> 21 months). Using banked tissue, Hartmann et al examined the expression of 30,721 genes using cDNA microarrays. Supervised learning algorithms were applied to a training set (n = 51) to develop a binary classifier, which was then applied to a testing cohort (n = 28) to assess predictive capacity. Fourteen genes comprised the final model. When applied to the testing cohort, outcome was successfully predicted in 86% of cases, with a 95% positive predictive value for early relapse. Survival characteristics between predictive cohorts (early vs late recurrence) were highly statistically significant—suggesting clinical relevance of the model. Hartmann et al conclude that validation of this model would allow for introduction of novel therapeutic regimens and strategies to patients with predicted poor outcome to standard primary therapy.
Comment by Robert L. Coleman, MD
The standard approach in the initial management of advanced ovarian cancer is cytoreductive surgery followed by combination chemotherapy. The most frequently used regimen in North America is taxane and platinum based, for which approximately 40-85% will achieve at least a partial response. Yet despite these high response characteristics, up to 80% of women will ultimately recur, usually within 2 years. To date, few models, other than clinical parameters (eg, stage, percent debulking, performance status, histology, CA-125, etc) exist to guide clinical expectations and as yet, the accuracy of these models is too imprecise to vary our initial treatment plan. Therefore, patients are generally treated as a class rather than individually.
Previous investigators over the past few years have looked at this issue mechanistically—evaluating genetic events that either help to define the molecular events leading to cancer or characterize response to primary therapy, such as chemo-resistance.1-3 What is novel about the current report is that the gene profiling is developed in a homogenous and representative cohort of advanced cases and done through supervised training of randomly sampled markers. This latter point is important, as superior predictive performance was observed over a whole data approach. The top 500 candidate markers (determined by signal-to-noise ratios) were then used in repetitive modeling in the training set. The 14 genes comprising the final model were then used in an independent evaluation set. The precision (86%) to predict outcome on either side of the median survival in the evaluation set (21 months) is remarkable. In particular, the ability to predict poor outcome (95%) is of great value, as these patients would have less than 1 in 20 chance of achieving median survival. The implication is that these patients, now identifiable, could be offered alternative therapy in the hopes of improving outcome.
This latter supposition is likely to be even trickier than the identification process. Over the last 15 years, only 1 agent has been identified that has favorably altered survival in primary therapy—paclitaxel. However, the explosion of active agents coupled with new insights into the biology of neoplasia offer some hope that we may be able to start the process of individualization of cancer therapy.
References
- Berchuck A, et al. Am J Obstet Gynecol. 2004;190:910-925.
- Lancaster JM, et al. J Soc Gynecol Investig. 2004;11:51-59.
- Wei SH, et al. Clin Cancer Res. 2002;8:2246-2252.
Subscribe Now for Access
You have reached your article limit for the month. We hope you found our articles both enjoyable and insightful. For information on new subscriptions, product trials, alternative billing arrangements or group and site discounts please call 800-688-2421. We look forward to having you as a long-term member of the Relias Media community.